汽機車之車牌辨識系統研究

一個車牌辨識系統，主要分為車牌定位、字元切割與字元辨識等三大部分。然而這三部分是互相影響，任何一部份之失真都將造成整體車牌辨識系統效率大幅降低。因此，在本論文中提出利用類神經網路之學習能力，使其能快速又正確地辨識出車牌號碼。本系統主要由定位模組及辨識模組所建構；前者先對影像做前處理再利用簡單之水平掃描，找出車牌上下邊界位置，再針對車牌影像做垂直掃描及字元切割，後者使用一個學習加權值之類神經網路來訓練學習並辨識車牌號，以提升該車牌辨識系統之辨識成功率。

關鍵字

車牌辨識系統；車牌定位

並列摘要

License Plate Recognition System (LPRS) divides into three parts: License Plate Location Finding (LPLF), Character String Segmentation, Enhancement (CSSE), and Character Recognition (CR). And they influence each other. Any part of them with distortion will severely decrease the efficiency of LPRS. In this paper, we use the learning ability of Artificial Neural Networks (ANN) to identify the number of car license plate quickly and correctly. We build the system consisted of orientation module and recognition module. The orientation module will pre-process the image, and scan the plate horizontally to find the boundary. Then scan the plate vertically and do the procedure of CSSE. The recognition module will use ANN of weighting value to train and learn to identify the license plate number to increase the percentage of recognition success.